Structure and Parameter Learning Driven Fault Diagnosis Method of Hydraulic Control System of Subsea Blowout Preventer

Hydraulic control employs compressed fluids as both the energy carrier and the conduit for information transmission. This approach has gained extensive application within industrial control systems due to its inherent adaptability and unwavering reliability. However, hydraulic systems are notorious for their high fault obscuration, substantial sensor latency issues, and intricate mechanisms for signal propagation. Consequently, pinpointing system malfunctions becomes exceedingly challenging when contending with the potent dynamics of nonlinear, time-variant characteristics at play. The diagnosis of complex hydraulic control system is a problem facing the current research. In order to cope with these challenges, a fault diagnosis method combining parameter learning and structure learning is proposed. A fault diagnosis model based on Bayesian networks of hydraulic control system is established, which realizes the diagnosis of common faults in hydraulic control system. The independence test based on Chi-square distribution was used to test the model and realize the structure learning function. The efficacy of the proposed method is exemplified through a case study involving a redundant control system designed for subsea blowout preventers, with the results affirmatively illustrating its high degree of accuracy.